USING PROBABILISTIC FORECASTS TO IMPROVE PLANNING AND OPERATIONS FOR WATER RESOURCES SYSTEMS W. JOSH WEISS, PH.D., P.E. jweiss@hazenandsawyer.com 2013 Susquehanna Water Science Forum October 7, 2013
OUTLINE What kinds of forecasts are available for water resources managers? What are probabilistic forecasts and why are they important? How do we use them for planning and operations? Example New York City s Operations Support Tool (OST) Forecast-based dynamic release program for the upper Delaware River Basin
RISK IN WATER RESOURCES PLANNING AND MANAGEMENT Flow in Esopus Creek at Allaben, NY Million gallons per day 2500 2000 1500 1000 500 We need to predict things that are variable and uncertain... 0 1/1/2004 2/20/2004 4/10/2004 5/30/2004 7/19/2004 9/7/2004
RISK IN WATER RESOURCES PLANNING AND MANAGEMENT Million gallons per day 1600 1500 1400 1300 1200 1100 1000 NYC In-City Demand... as if they are fixed and certain. 900 800 1/1/2011 2/1/2011 3/1/2011 4/1/2011 5/1/2011 6/1/2011 7/1/2011 8/1/2011 9/1/2011
RISK IN WATER RESOURCES PLANNING AND MANAGEMENT We can t avoid risks or tradeoffs among competing objectives But we can improve decisions by better understanding uncertainty and risks Forecasts and system models offer a framework Source: New York Times
SUPPLY AND DEMAND FORECASTING IS NOT A PERFECT SCIENCE Supply Forecasting Uncertainties Near-term climate variability Long-term climate non-stationarity Land use changes Stream geomorphological characteristics Depletion of existing sources Infrastructure Regulations Demand Forecasting Uncertainties Future climate Future growth patterns Population dynamics Demographics Economics Per capita water use
WE SEEK TO EXPLAIN UNCERTAINTIES AND EXPLICITLY ACCOUNT FOR THEM Possible departures from anticipated socioeconomic conditions, longterm normal weather Resulting uncertainty in forecasted regional demand
TYPES OF FORECASTS Deterministic Historical analogs Mechanistic / physical models Probabilistic / stochastic / ensemble Statistical analysis of historical analogs Mechanistic / physical with uncertainties
Annual Average Daily Supply or Demand, mgd DETERMINISTIC FORECASTS: ONE PREDICTION OF THE FUTURE 240 Future Need 210 Future Supply 180 150 Future Demand 120 2015 2005 2010 2020 2015 2025 2020 2030 2025 2035
Annual Average Daily Supply or Demand, mgd PROBABILISTIC FORECASTS: RANGE OF FUTURE PREDICTIONS 240 Future Need 210 Future Supply 180 150 Future Demand 120 2015 2005 2010 2020 2015 2025 2020 2030 2025 2035
PROBABILISTIC FORECASTS: RANGE OF FUTURE PREDICTIONS
PROBABILISTIC FORECASTS: RANGE OF FUTURE PREDICTIONS Observed Predicted
PROBABILISTIC FORECASTS: RANGE OF FUTURE PREDICTIONS 50 th 75 th 25 th Observed Predicted
PROBABILISTIC FORECASTS: RANGE OF FUTURE PREDICTIONS 50% of predictions indicate reservoir does not refill Observed Predicted
SYSTEM MODELING TOOLS
SYSTEM MODELING TOOLS: INPUTS Inflow Forecasts Demand Forecasts System Infrastructure System Operating Rules Water Quality Models
SYSTEM MODELING TOOLS: OUTPUTS Storage Projections Shortfall Projections Stream Flows at Key Locations Probability of Drought Status Water Quality Impacts
NEW YORK CITY S OPERATIONS SUPPORT TOOL (OST)
NEW YORK CITY S OPERATIONS SUPPORT TOOL (OST)
DELAWARE RIVER BASIN Headwaters in the Catskill Mountains Headwaters impounded by three NYC water supply reservoirs Drains to New York State, Pennsylvania, New Jersey, and Delaware Water source for NYC, New Jersey (D&R Canal), Philadelphia Ecological and recreational value Salinity repulsion at the Delaware Bay
DELAWARE RIVER BASIN INTERSTATE WATER RESOURCES MANAGEMENT 1954 U.S. Supreme Court Decree 1961 Delaware River Basin Compact 1961-1967 New drought of record 1983 Good Faith Agreements 2008 Flexible Flow Management Plan Images from DRBC website: www.state.nj.us/drbc/
2008 FLEXIBLE FLOW MANAGEMENT PLAN (FFMP) Relied on release schedules based on annual estimation of available water Expired on May 31, 2011 Opportunity to implement better approach to estimating available water, using near-term ensemble hydrologic forecasts
OST FORECAST-BASED RELEASE PROGRAM FOR NYC RESERVOIRS Use bi-weekly Forecasted Available Water (FAW) calculations to determine release quantity at a specified risk level
FORECASTED AVAILABLE WATER BALANCE Today s Total PCN Storage Current System Status + Cumulative PCN Inflows through June 1 Cumulative PCN Diversions through June 1 Probabilistic Streamflow Forecasts Estimated Volume to meet NYC Demand June 1 Storage Target 100% Usable Storage = Cumulative PCN Release Target through June 1 Distribute over Number of Days to June 1 and Re- Evaluate Regularly
RISK-BASED APPROACH Risk represented by forecast probability Transition to more conservative risk when approaching June 1 Seasonal risk factor: Forecast Percentile 80 70 60 50 40 30 20 10 0 70% probability of less inflow 50% probability of less inflow 30% probability of less inflow Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May
ENHANCED FLOOD MITIGATION RELEASES Conditional Storage Objective (CSO) for enhanced flood mitigation rule
ENHANCED FLOOD MITIGATION MASS BALANCE Today s Total PCN Storage Current System Status + = Cumulative PCN Inflows over the Next 7 Days Cumulative PCN Release over the Next 7 Days Cumulative PCN Diversions over the Next 7 Days Conditional Storage Objective (CSO) Predicted Storage Surplus Relative to CSO Streamflow Forecasts (50 th Percentile) Based on OST-FFMP Table Selection Estimated Volume to Meet NYC Demand Boundary between L1-b and L1-c Zones Release Estimated Surplus; Re-Evaluate Daily
RULE TESTING: LONG-TERM SIMULATION WITH HINDCASTS One 80-year simulation
Neversink Release PCN Release Cannonsville Spill Pepacton Spill RULE TESTING: PCN Spill LONG-TERM SIMULATION WITH HINDCASTS 0 100 200 300 400 500 600 700 800 900 Neversink Spill Cannonsville Release Pepacton Release Neversink Release PCN Release Cannonsville Spill Pepacton Spill Neversink Spill PCN Spill Average Flow (cfs) Percent Change in Average Simulated PCN Release and Spill -40% -30% -20% -10% 0% 10% 20% 30% 40% Percent Change (OST-FFMP relative to FFMP-35)
IMPLEMENTATION: REAL-TIME POSITION ANALYSIS MODE
IMPLEMENTATION: REAL-TIME POSITION ANALYSIS MODE
BASIN-WIDE ADOPTION CHALLENGES How to bring stakeholders on board? How to minimize black box perceptions? How to ensure transparency while protecting NYC system information? How to overcome historical mistrust?
POTENTIAL FUTURE DEVELOPMENT: BASIN-WIDE MANAGEMENT TOOL Existing OASIS model for long-term simulations Existing database framework and ensemble forecasts throughout the Basin Give stakeholders a realtime tool to support negotiations and reduce black-box perceptions
WHAT IF THE FORECAST IS WRONG?
CLIMATE CHANGE VULNERABILITY ASSESSMENT AND ADAPTATION PLANNING Assessment of climate variability impacts & adaptation options OST used in two recent Water Research Foundation projects Decision-making and Planning under Uncertainty (#4262) Dynamic Reservoir Operations (#4306)
FORECAST-BASED APPROACH: Represents a conceptual shift to explicit, quantitative consideration of risks and tradeoffs Can facilitate a better understanding of risks and vulnerabilities Can help managers prioritize objectives, make better use of a limited resource Framework for developing win-win solutions jweiss@hazenandsawyer.com